o .wi @sddlZddlmZddlZddlmZddlmZddlmZddl m Z m Z m Z dedefd d Z dd ededededdeedef ddZdedefddZ dd edededdeedef ddZ ddededdeedefddZdS)N)Optional)Tensor)Literal)#_multiclass_confusion_matrix_update)_drop_empty_rows_and_cols_handle_nan_in_data_nominal_input_validationconfmatreturncCsVt|}|}||}|d|}|dd|jd}t|t||S)aCompute Conditional Entropy Statistic based on a pre-computed confusion matrix. .. math:: H(X|Y) = \sum_{x, y ~ (X, Y)} p(x, y)\frac{p(y)}{p(x, y)} Args: confmat: Confusion matrix for observed data Returns: Conditional Entropy Value )rsum unsqueezerepeatshapetorchnansumlog)r total_occurrencesp_xy_mp_yp_y_mre/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/torchmetrics/functional/nominal/theils_u.py_conditional_entropy_computes  rreplacepredstarget num_classes nan_strategy)rdropnan_replace_valuecCsN|jdkr |dn|}|jdkr|dn|}t||||\}}t|||S)aCompute the bins to update the confusion matrix with for Theil's U calculation. Args: preds: 1D or 2D tensor of categorical (nominal) data target: 1D or 2D tensor of categorical (nominal) data num_classes: Integer specifying the number of classes nan_strategy: Indication of whether to replace or drop ``NaN`` values nan_replace_value: Value to replace ``NaN`s when ``nan_strategy = 'replace``` Returns: Non-reduced confusion matrix r )ndimargmaxrr)rrrrr!rrr_theils_u_update7s r%cCs`t|}t|}|}|d|}t|t| }|dkr*tjd|jdS|||S)zCompute Theil's U statistic based on a pre-computed confusion matrix. Args: confmat: Confusion matrix for observed data Returns: Theil's U statistic rdevice)rrr rrtensorr')r s_xyrp_xs_xrrr_theils_u_computeQs  r,cCs.tt||g}t|||||}t|S)aGCompute `Theils Uncertainty coefficient`_ statistic measuring the association between two nominal data series. .. math:: U(X|Y) = \frac{H(X) - H(X|Y)}{H(X)} where :math:`H(X)` is entropy of variable :math:`X` while :math:`H(X|Y)` is the conditional entropy of :math:`X` given :math:`Y`. Theils's U is an asymmetric coefficient, i.e. :math:`TheilsU(preds, target) \neq TheilsU(target, preds)`. The output values lies in [0, 1]. 0 means y has no information about x while value 1 means y has complete information about x. Args: preds: 1D or 2D tensor of categorical (nominal) data - 1D shape: (batch_size,) - 2D shape: (batch_size, num_classes) target: 1D or 2D tensor of categorical (nominal) data - 1D shape: (batch_size,) - 2D shape: (batch_size, num_classes) nan_strategy: Indication of whether to replace or drop ``NaN`` values nan_replace_value: Value to replace ``NaN``s when ``nan_strategy = 'replace'`` Returns: Tensor containing Theil's U statistic Example: >>> from torch import randint >>> from torchmetrics.functional.nominal import theils_u >>> preds = randint(10, (10,)) >>> target = randint(10, (10,)) >>> theils_u(preds, target) tensor(0.8530) )lenrcatuniquer%r,)rrrr!rr rrrtheils_uls)r0matrixc Cst|||jd}tj|||jd}tt|dD]9\}}|dd|f|dd|f}}tt ||g } t ||| ||} t | |||f<t | j |||f<q|S)aCompute `Theil's U`_ statistic between a set of multiple variables. This can serve as a convenient tool to compute Theil's U statistic for analyses of correlation between categorical variables in your dataset. Args: matrix: A tensor of categorical (nominal) data, where: - rows represent a number of data points - columns represent a number of categorical (nominal) features nan_strategy: Indication of whether to replace or drop ``NaN`` values nan_replace_value: Value to replace ``NaN``s when ``nan_strategy = 'replace'`` Returns: Theil's U statistic for a dataset of categorical variables Example: >>> from torch import randint >>> from torchmetrics.functional.nominal import theils_u_matrix >>> matrix = randint(0, 4, (200, 5)) >>> theils_u_matrix(matrix) tensor([[1.0000, 0.0202, 0.0142, 0.0196, 0.0353], [0.0202, 1.0000, 0.0070, 0.0136, 0.0065], [0.0143, 0.0070, 1.0000, 0.0125, 0.0206], [0.0198, 0.0137, 0.0125, 1.0000, 0.0312], [0.0352, 0.0065, 0.0204, 0.0308, 1.0000]]) r r&r"N)rrronesr' itertools combinationsranger-r.r/r%r,T) r1rr! num_variablestheils_u_matrix_valueijxyrr rrrtheils_u_matrixs "r=)rr)r3typingrrrtyping_extensionsr7torchmetrics.functional.classification.confusion_matrixr%torchmetrics.functional.nominal.utilsrrrrintfloatr%r,r0r=rrrrs`      0